Cameron Herold explores ElevenLabs' text-to-speech platform, which turns EPUBs into audiobooks and clones voices without studio time. He spent 2.5 days and $6,000 per book recording audiobooks; the platform could do it in days using his own voice trained on existing recordings.
Chapters
So talk to us a little bit about what ElevenLabs is working on.
ElevenLabs, as you mentioned, is a text-to-speech company. Our mission is to make content universally accessible in any language and voice, so you could dub a podcast that is still spoken in your voice into another language. That's like the goal. The more immediate use cases are inputting text and outputting audio. For example, if you want to create an audiobook, You can upload the EPUB file, that is the format in which audiobooks come, and then you can immediately instantly create the spoken version of that. You can even go as far as instead of having one voice read the entire audiobook, you can give each character their own voice and then have the dialogues with different characters, which I think is such a fun improvement to what, how we know audiobooks at the moment. So those are 2 examples of how text-to-speech can be used, and ElevenLabs is one of the leading AI companies in that space with the best models that are really great at creating emotionality in the voices and having contextual awareness as well, and this contextual awareness is where AI really, really feeds in and where AI has allowed a breakthrough for text-to-speech compared to how, you know, Alexa speaks or Siri speaks, where the model understands what it is saying and therefore can say it in exactly the right way. So you could say a sentence, it's like, hi, my name is Victoria, I'm really sad. And that will come out sounding a lot sadder than if you would say, hi, I'm Victoria, I'm so happy. That would come out like more excited because of the context of that sentence. And you could also clone voices so you could have that sentence spoken by me without me having to sit and record it in the first place in the happy and the sad version potentially.
Yeah, what I love about Eleven, and I've heard about Eleven now over the last few months almost weekly, someone has mentioned it to me for a few reasons, and I'll talk about why I'm excited and I'm going to become a client quite quickly. I love also that it can also add the inflection and the intonation into the speech as well so that it literally does take the written word and it doesn't sound robotic, it really does sound like we would speak. So the reason I'm excited about it, I went into studio a year ago today to record my 6th book. It's called The Second in Command. And I spent 2 and a half days in studio in Dubai recording my audiobook. And I literally felt like poking my effing eyes out with my pen. Like I was just like, someone kill me. This was 8 hours a day standing, trying to actually be excited when my content's really good, but it's hard for me to be excited reading it. I'm also not that great at reading. I'd rather stand in front of 1,000 people than stand and read to someone. And then to try to throw the inflection and the intonation in and without having to like read it and then also go on these little hyperlinks because I wanted to add more information that was probably coming in the next chapter anyway. And then all of a sudden I find out that I could literally now, like now that I've read that entire book, I can upload it. I can take a few of my speaking events and upload that content that will train your system AI. And then I could upload my other 5 books and your system could literally punch out the audio recording for me without me having to spend what would now be 14 more days in studio.
Exactly. You could be done today.
It's fucking amazing. It also, when I had to pay the authors to read my other 5 books, I paid them $6,000 each. I don't love their voice. They do sound a little bit kind of robotic. And strangely, people still want, because I've done so many paid speaking events, I've been paid to speak in 28 countries. I have this audience of people that want to read my content, but they want to hear me reading it for some reason. Like we want to hear—
And they want your voice.
Yeah. And they, they just don't like Fred who's reading it, even though Fred sounds fine. Right. So how the fuck does the system work? How does this, how does Eleven work?
Another use case, also quickly before we dive in a bit more into the tech, the other day when you lost your voice, you potentially could have continued recording podcasts if you would've, uh, uh, implemented some text-to-speech there as an idea. So, uh, both saving time in the studio, but also Saving time at home.
I did, I did actually also write down that I'm going to take a look at using ElevenLabs to also transcribe our podcast into a number of the best languages. Like, there's no reason why we shouldn't do, you know, Mandarin, Spanish, French, German. That's probably enough. I mean, like, what would the big 5 or 6 be?
Those are definitely on the top of the list. Definitely some that we, that we, that ElevenLabs offers as well. ElevenLabs has 28 languages that you can choose from, many, many more voices. But you can also upload your own voice and Mandarin, Spanish, German, French are all in there. You could even do English to English from your voice to Fred's voice or vice versa.
